I've had more time to dip my toes in fully -- I had ZiPS to finish and I'm working on some positional WAR adjustments -- and with a few experiments, I'm cautiously feeling @tangotiger.com is on to something that the less linearity of WAR $ suggests WAR is set too low. - ThreadSky

dszymborski.fangraphs.com • 34 days ago

I've had more time to dip my toes in fully -- I had ZiPS to finish and I'm working on some positional WAR adjustments -- and with a few experiments, I'm cautiously feeling @tangotiger.com is on to something that the less linearity of WAR $ suggests WAR is set too low.

Comments

nepstein10.bsky.social•34 days ago

How much of this might be attributable to teams being willing to sacrifice some W's for some $s? We know some teams definitely do this in general (*cough* Pirates, Marlins, etc.), but is there a general trend of taking a lower value player making the league minimum over a vet coming into play?

derkomminsksar.bsky.social•34 days ago

Not that you asked but - I agree about the replacement level having been set a little too low for position players. For pitchers, I’d be curious as to how people think changing practice patterns (in a world with roster limits) impacts these calculations.

war2d2.bsky.social•34 days ago

I must have missed this, do you have a link where he wrote about it? When you say “too low” do you mean the distribution should favor the “right tail” of the bell curve more than it does, or that there should be a bigger pool of WAR to distribute?

war2d2.bsky.social•34 days ago

I read through the thread again and see you meant replacement has been set too low, sorry about the dumb question.

dszymborski.fangraphs.com•34 days ago

Which does make sense from an overall standpoint. Our pinpointing of replacement level was largely based on how teams *acted* and teams are far more competent at evaluating their talent than they were 20 years ago.

dszymborski.fangraphs.com•34 days ago

(My apologies to @tangotiger.com for not really addressing the idea last month, I didn't want to say anything until I had at least *some* substantive opinion to offer)

dszymborski.fangraphs.com•34 days ago

ZiPS has a model for "taking away" a player's playing time in the long horizon based on performance, age, contract. But there's a definite implication in the model that teams are willing to replace players at significantly higher levels of competence than they used to.

dszymborski.fangraphs.com•34 days ago

I really need to dig into more to make a truly cogent argument, but the differences are stark pretty suddenly around 2010. I think there's a real chance that the "true" (a difficult concept) replacement level should have always been a bit higher than it is.

dszymborski.fangraphs.com•34 days ago

There is no solidity here as I haven't really sat down and done a robust look, but I'm seeing a real implication from team behavior that replacement level is *probably* best set in the low-mid 50s in wins. There are no doubt a lot of philosophical implications.

dszymborski.fangraphs.com•34 days ago

Remember, WAR is a framework, that evolves as we have more knowledge. I think there's at least a decent argument at the extremes, that WAR's replacement level *is* at least moderately overrating compilers and underrating peak guys, because we based on very oldschool team behavior

Comments

Posting Rules

Reply